Author: IRPA AI Senior Analyst, Kieran Gilmurray
Most AI conversations begin with the wrong question. Leaders ask what can be automated, accelerated, delegated to agents, or supported by AI, but the better first question is more uncomfortable: should this work still exist in its current form?
This article explores why strategic subtraction is becoming a critical leadership discipline in the AI era. If organisations use AI to accelerate obsolete reports, duplicated approvals, overloaded meetings, fragmented tools, or legacy governance rituals, they do not create transformation. They create faster clutter.
The Wrong First Question
Most organisations start AI transformation by asking what can be automated. That sounds sensible, but it can be strategically dangerous. If the starting point is automation, the organisation often assumes the existing workflow deserves to survive. The only question becomes how to make it faster.
That is how companies end up automating clutter. A report that nobody uses becomes easier to produce. A duplicated approval becomes faster to complete. A meeting generates a better summary but still does not lead to a decision. The organisation creates more output, but not necessarily more value.
The better first question is not “Can AI do this?” It is “Should this still be done?” That question changes the conversation from automation to redesign, and from productivity theatre to real operating value.
Why AI Can Make Clutter Worse
AI changes the economics of production. Drafting, summarising, searching, coding, reporting, and analysis can all become faster, but faster production does not automatically create better performance. In many organisations, the constraint is no longer the ability to produce more information. The constraint is the ability to absorb it.
Leaders are already dealing with too many reports, meetings, dashboards, tools, updates, approvals, notifications, and competing priorities. AI can add even more volume into a system that is already overloaded. Microsoft finds that organisational conditions explain more than twice the reported AI impact of individual behaviour, PwC reports that 20% of firms capture 74% of AI’s economic value, and the World Economic Forum reports that only about 15% of organisations are using AI to fundamentally redesign how work is performed.
The message is clear. AI value does not come from accelerating everything. It comes from redesigning the system around what matters, removing what no longer belongs, and protecting the work that genuinely creates trust, value, resilience, and judgement.
Strategic Subtraction Is Not Cost Cutting
Strategic subtraction is the deliberate removal or redesign of organisational activity that no longer creates enough strategic value relative to the friction, duplication, risk, and capacity it consumes. It focuses on work, not just cost.
That distinction matters. Cost cutting asks, “Where can we spend less?” Strategic subtraction asks, “What should no longer occupy attention, coordination, governance, and decision capacity?” It can apply to meetings, status reports, approvals, dashboards, handoffs, governance steps, tools, workarounds, initiatives, queues, and legacy routines.
Some work should be removed entirely. Some should be simplified. Some should be consolidated. Some should be paused. Some should be hidden from default workflows but retained for specialist use. Some should be protected because it still carries trust, risk, learning, or resilience value. This is why strategic subtraction is not crude reduction. It is disciplined judgement.
The VITALS Test
The Strategic Subtraction Test is built around six questions. Leaders can use it before approving new automation, agents, AI workflows, transformation activity, or operating model change.
Value: Does this activity directly improve revenue, customer outcomes, decision quality, risk reduction, strategic learning, or operational performance?
Interference: How much delay, rework, queueing, review burden, search, switching, meeting load, or coordination does it create?
Twins: Is the same information, approval, control, report, dashboard, meeting, or initiative duplicated somewhere else?
Assurance risk: If changed or removed, would trust, resilience, compliance, safety, customer promise, or learning capacity weaken?
Liberation of capacity: What time, budget, leadership attention, or decision bandwidth would be released, and is that capacity actually reclaimable?
Strategic fit: Does this activity support the current strategy and future operating model, or is it a legacy artefact from a previous constraint?
The test works best when applied to real work objects: a meeting series, dashboard, approval step, report, review queue, governance forum, workflow, tool, or initiative. A strong subtraction candidate usually has low Value, low Strategic fit, high Interference, high duplication, low Assurance risk, and meaningful capacity release. If Assurance risk is high, the answer is rarely simple removal. It is more likely protection, redesign, or risk tiering.
How to Use the Test
The easiest way to use VITALS is to start with one workflow where AI adoption is already happening or where leaders are under pressure to improve speed, cost, quality, or value. Do not begin with the official process map. Begin with how the work actually moves through the organisation.
Map the meetings, reports, dashboards, approvals, tools, handoffs, review queues, escalation steps, and manual workarounds. Then score each one against VITALS. The aim is not to prove that everything should disappear. The aim is to separate what should be eliminated, simplified, consolidated, paused, hidden, substituted, or protected.
For example, a weekly deck might be substituted with a live workflow view. A low risk sequential approval chain might be simplified into risk tiered review. Two overlapping governance forums might be consolidated. A report might be hidden from default circulation but retained for audit or specialist use. A customer complaints review might be protected because it creates learning and trust, even if it feels slow.
The discipline is to make the action explicit. Most subtraction fails because new work is added but old work is never formally retired.
What Leaders Should Remove First
The best starting point is usually not headcount. It is work. Leaders should look first at standing meetings with no decision, reports that do not change action, duplicated dashboards, overlapping governance forums, low risk approvals that move through too many hands, tools that create search and reconciliation work, and legacy initiatives that continue because nobody formally stopped them.
These are attractive subtraction candidates because they often consume attention without creating proportional value. They also create drag around AI adoption. If the old meeting, report, approval, or dashboard remains in place, AI simply becomes another layer on top of the old system.
Atlassian’s evidence on software delivery is a useful warning. AI may save coding time, but developers still lose significant hours to non coding work such as fragmented tools, unclear requirements, meetings, and information search. The lesson travels beyond engineering. If leaders automate the fastest part of the workflow but leave the surrounding friction untouched, the value leaks elsewhere.
What Leaders Should Protect
Strategic subtraction only works when leaders know what must not be removed. Some work looks inefficient because its value is protective rather than obvious. Legal controls, cyber checks, privacy safeguards, safety processes, customer complaint signals, incident reviews, escalation routes, override rights, and junior learning loops may all appear slow in a narrow productivity review.
Removing them without redesign can damage trust, resilience, compliance, capability, and judgement. That is why Assurance risk is central to the test. Leaders should not ask only whether something is slow. They should ask what it protects.
A duplicated approval may be waste. A human review in a high consequence decision may be trust infrastructure. A routine report may be stale. An incident review may be a learning loop that protects the organisation from repeating mistakes. Good subtraction removes clutter while protecting capability.
The Common Failure Pattern
The most common failure pattern is adding AI without retiring old work. A team introduces an AI assistant, but the old meeting continues. A workflow becomes faster, but the same approvals remain. Reports become easier to generate, so more reports appear. A new agent handles first pass analysis, but review queues grow because nobody redesigned the escalation model.
That is how AI creates work intensification rather than capacity release. Berkeley Haas found that AI widened job scope, increased work density, and dissolved stopping points in the working day. BCG’s research on “AI brain fry” similarly warns that excessive AI use or monitoring can increase decision overload, errors, and intent to quit.
If AI increases output but leaders do not subtract work around it, the system absorbs the gain as more activity. The organisation becomes busier, not better.
What This Means for Leaders
Leaders should make strategic subtraction a formal part of AI governance, transformation planning, and operating model design. Before approving new automation, agents, copilots, or transformation workstreams, they should ask which existing work will be removed, simplified, consolidated, paused, hidden, substituted, or protected.
This is not a side exercise. It is central to value realisation. AI will not deliver its full impact if it is layered onto overloaded workflows, duplicated controls, fragmented tools, and legacy routines. The organisation has to create capacity before it can absorb new capability.
The strongest leaders will not ask only what AI can do. They will ask what the organisation should stop carrying forward. In the AI era, removing the wrong work may become just as important as accelerating the right work.
About the Author: Kieran Gilmurray
Kieran is a globally recognized authority on AI, automation, and digital transformation, having authored multiple influential books and hundreds of articles that have earned him prestigious accolades, including being named a Top 50 Global Thought Leader and Influencer on Generative AI in 2024, a Best LinkedIn Influencer for AI and Marketing, Top 50 Global Thought Leaders and Influencers on Manufacturing 2024, Top 14 people to follow in data and one of the World’s Top 200 Business and Technology Innovators.
CLICK HERE TO SCHEDULE AN ANALYST CHAT
Links:
Originally posted on 2025-01-28 in the IRPA AI Network — Announcements & Updates